The FIM section is continuing an effort to develop more robust, informative, and quantitative methods for mapping human brain function in both the activated state and during resting state. We are focusing on processing methods, brain activation intensity and timing manipulations, and physiologic manipulations to better understand the fMRI signal change and sources of artifact. To summarize, the following were among the highlights, organized by research project corresponding to the listed staff scientist, student, or post doc. Rasmus Birn: The effect of task-correlated physiological noise: Variations in the subjects heart rate and breathing pattern can result in significant fMRI signal changes, mediated in part by non-neuronal physiological mechanisms such as global changes in levels of arterial CO2. This is a concern for studies involving emotional stimuli or changes in arousal. In this study, we showed that physiological noise can be separated from neuronal-induced BOLD responses by exploiting the differences in the spatial and temporal profiles of these two types of signal changes. More specifically, the delineation of activation in the presence of task correlated breathing changes was improved either by independent component analysis, or by including specific nuisance regressors in a regression analysis. The success of these strategies suggests that activation can be studied even in the presence of task-correlated physiological changes. Comparison and Validation of fMRI Calibration Techniques: The amplitude of the blood oxygenation level dependent (BOLD) fMRI signal depends strongly on the properties of the underlying vasculature, particularly the baseline blood volume. In this study we test various techniques that have been proposed for calibrating the amplitude of the BOLD response across the brain, and compare their ability to pull out subtle variations in neuronal activity. We find that spatial calibration of the BOLD signal using the response to breathing variations or the temporal variance at rest improves the ability to see subtle variations in neuronal activity. Niko Kriegeskorte: FMRI voxel as compact-kernel or complex spatiotemporal filter: In the past year, we finished a theoretical and simulation project exploring the possible physiological basis of the presence of sub-voxel-scale neuronal pattern information in fMRI response patterns. Our results suggest that voxels are best viewed as complex spatiotemporal filters, where each voxel may have a unique filter sensitive to fine-grained pattern information. Human object similarity judgments: We developed a novel method for acquiring large sets of pairwise similarity judgments from human subjects. Subjects arrange the objects on a computer screen by mouse drag and drop. A sequence of trials, in which subsets of the object set are arranged by the subject, allows us to estimate a high-dimensional subjective similarity structure for large sets of objects (e.g. 100 objects), where the number of pairs is too large ((1002-100)/2=4950) for separate pairwise ratings. We validated the method by comparison to conventional methods and compared the human similarity judgments for 96 object to the similarity structure in various brain regions. Results indicate that IT best predicts human object dissimilarity judgments (among early visual and ventral stream regions). Face-identity change elicits widespread activation in the visual system: We found that subsequent presentation of faces of different people leads to greater activation than face repetition in number of brain regions not specialized for face processing including early visual cortex. This suggests that the common interpretation of such effects in the Fusiform Face Area (FFA) as indicating an individual-face representation (release from adaptation) may be mistaken: more general attentional factors may explain these and other release from adaptation fMRI effects. Masaya Misaki: Comparing classification analysis methods in fMRI pattern information analysis: Six multivariate classification methods were compared with regard to their abilities to extract information in fMRI activation patterns and robustness to various types of noise in fMRI data. Due to the large dimensionality of the data relative to the sample size, linear methods had better performance in the fMRI pattern-information analysis. Extracting face-category and face-exemplar information in the IT cortex from its spatial and spatio-temporal response patterns: Face-category and the face-exemplar information was extracted using multivariate decoding methods. The spatial pattern and the spatio-temporal pattern of the BOLD response were used as inputs for classification analysis to decode stimulus category (face or house) and face identification. For decoding face category, spatial pattern contained most information. For decoding face exemplar, spatio-temporal pattern had more information than the spatial pattern. Decoding a sub-second (500 ms) temporal difference of visual stimulus presentations in the early visual cortex from the BOLD response: To investigate whether subtle temporal processing differences may be extracted using BOLD contrast, decoding performance was examined with a task in which temporal differences in visual stimulus presentations across hemifield was modulated (1000 and 500 ms difference). Spatio-temporal patterns of the BOLD response was used for decoding of temporal order. A 1000ms difference was decoded reliably for all subjects, and 500ms difference was decoded for half of the subjects. Qu Tian: Acute effects of exercise on fMRI signal, perfusion, and anatomic measures: She was a summer student who had a project that focused on the effects of elevated physical activity on the fMRI signal change resting state fluctuations, activation dynamics, perfusion, and anatomy. So far we have been detecting trends in which post-exercise the fMRI signal changes have been somewhat reduced. We are in the proces of further analyzing our perfusion, resting state fluctuation, and anatomic data. Paula Wu: Effects of the valsalva maneuver on fMRI signal: She was a summer student who had a project that focused on the detailed dynamics of the valsalva maneuver effects on fMRI signal. One calibration method for fMRI is the use of a breathold to increase blood flow globally. This is sometimes performed - improperly - in combination with the valsalva maneuver (a closing of air passages and applying pressure). We have determined that this effect on the MRI signal is profound but may have utility as it may reflect rapid blood volume changes. Dan Handwerker: Voxel-wise Correlation of resting state fluctuations with task and global stress magnitude: We examined the magnitude of signal fluctuations in resting fMRI scans and determined the correlation with magnitude changes in response to tasks and breath holding. The resting magnitudes showed a strong correlation with task- based magnitudes even after known respiration and pulsatory noise was removed. Proof of artifactual anti-correlated networks: We examined what removing the global signal does to resting-state correlations with fMRI. The larger the correlation before removal of the global signal, the more the correlation shifts towards zero after the global signal is removed. This study further proves that the creation of anti-correlated networks is a mathematical artifact of removing the global signal. Examination of non-stationarity of resting state correlations: We examined how resting-state correlations change over time using a small sliding correlation window. We show that pairs of interacting regions reveal novel network properties. This opens up a new dimension in the study of resting state correlations
Showing the most recent 10 out of 84 publications
